tandem.cpp: Qwen3.8-Flash-Next at 97 tok/s on a single 4090 + 64 GB RAM, 3.3× llama.cpp, works for any MoE

🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xldent/tandemcpp_qwen38flashnext_at_97_toks_on_a_single/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-11-02 04:23:50
Reddit
From the LocalLLaMA community on Reddit
From the LocalLLaMA community on Reddit
Explore this post and more from the LocalLLaMA community